LM334 questions, and it's role in this schematic.

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

AVA

Well-known member
Joined
Jun 7, 2021
Messages
47
Location
Canada
Hi everyone, I hope you're all doing well.

I included a snippet of a schematic that came from here: THATCorp 4315 One Knob Squeezer PDF Link

It's a compressor circuit, and the snippet I provided is where the control voltage gets manipulated.

In my (tiny) brain, the LM334 is basically just an amazing resistor right? It 'sources current' and does it with incredible accuracy so...its an incredible resistor lol right? Why do I feel like I'm gonna get eaten alive for this sentence..

Anyway, in this circuit, very near to the 334 there is a "+VP" and "-VP" as well as an inverting op amp (U5A)...

- Is this small portion of the circuit actually creating a dual supply for the "more" potentiometer VR1? The '+VP' is also used for VR2, but nowhere else. I guess I'm a little confused what VP stands for lol.

- If I'm right (for the first time), and the whole point of this little 334 and surrounding circuitry is to create a dual supply for the purpose of the 'more' control, why not just use a charge pump at the power supply and run everything off of it? It would avoid having to DC bias anything, eliminate the voltage divider and buffer at the power supply...

- Is it because charge pumps (1044, for example) don't supply enough current? Or are unreliable?
The LM2662 could put out up to 200mA, but is 5.5V Max. Still, a 10V span for a small audio signal is quite sufficient...

Perhaps the THAT4315 wouldn't enjoy a dual supply, but for a similar price, the 4305 can. Hmm...


Thanks in advanced! Also, apologies in advanced for any toes I might have somehow stepped on without realizing lol.
 

Attachments

  • LM334.png
    LM334.png
    273.8 KB · Views: 78
  • Zoomed Out.png
    Zoomed Out.png
    365.2 KB · Views: 73
Perhaps I should explain the impetus behind my recent obsession with THATCorp circuits lately...

I'm basically just building all their circuits on a breadboard, tinkering around, swapping things, trying stuff, and absorbing the lessons.

I think there's a lot to learn from their circuits, and they have a lot of proven schematics to offer, so the pool is deep.

I didn't have the pleasure of going to school for this, back when I could afford the time and money to do it. So aside from reading all the books recommended on this forum, this is how I get my education... breadboard - tinker - ponder - question - make notes - rinse and repeat.

Sometimes it means I have to ask a dumb question or two. So apologies for that, and thank you for your patience.

My current goal is to create a simple soft knee compressor, ideally using the THAT4315 on a single 9V supply.


Anyway, cheers! Excited to see what comes up about the LM334
 
- Is this small portion of the circuit actually creating a dual supply for the "more" potentiometer VR1? The '+VP' is also used for VR2, but nowhere else. I guess I'm a little confused what VP stands for lol.
It is often convenient to name a node specifically. I see nothing wrong in naming this precise voltage rail, a it could be used later in other variants of the circuit
- If I'm right (for the first time), and the whole point of this little 334 and surrounding circuitry is to create a dual supply for the purpose of the 'more' control, why not just use a charge pump at the power supply and run everything off of it? It would avoid having to DC bias anything, eliminate the voltage divider and buffer at the power supply...

- Is it because charge pumps (1044, for example) don't supply enough current? Or are unreliable?
Charge pumps are often limited and they are noisy, not only at their output, but also the pollute rails if nor heavily by-passed.
The LM334 solution is very accurate and extremely quiet.
The LM2662 could put out up to 200mA, but is 5.5V Max. Still, a 10V span for a small audio signal is quite sufficient...
That means one would need to drop the battery voltage from 9V to 5, which would either waste a lot of energy or require a switching regulator, that would add even more rail pollution.
 
I think there's a lot to learn from their circuits, and they have a lot of proven schematics to offer, so the pool is deep.
There's many brilliant minds at THAT.
I didn't have the pleasure of going to school for this, back when I could afford the time and money to do it. So aside from reading all the books recommended on this forum, this is how I get my education... breadboard - tinker - ponder - question - make notes - rinse and repeat.
Good luck with it!
 
It is often convenient to name a node specifically. I see nothing wrong in naming this precise voltage rail, a it could be used later in other variants of the circuit

Charge pumps are often limited and they are noisy, not only at their output, but also the pollute rails if nor heavily by-passed.
The LM334 solution is very accurate and extremely quiet.

That means one would need to drop the battery voltage from 9V to 5, which would either waste a lot of energy or require a switching regulator, that would add even more rail pollution.
Hey, thanks for your time. I appreciate your reply

I agree, they definitely have some brilliant people working there. After building so many of their circuits I’m really inspired by them.

Some of their resistor values are interesting tho lol Like the soft knee compressor…it’s amazing, but, like, 24k3? 1M43? 12k4? 1k91? Haha that’s not a criticism, I love it.

By your response I’m assuming that it is in fact creating a dual supply for that portion of the circuit. Which is cool because that means I’m onto something here lol

But why the 334? I mean, why not just have this exact circuit but without the 334? Is it because the RMS portion of the 4315 will consume spikes of high current in proportion to the transients that are fed into it?
 
Some of their resistor values are interesting tho lol Like the soft knee compressor…it’s amazing, but, like, 24k3? 1M43? 12k4? 1k91?
Typical values for 1% tolerance resistors.
But why the 334? I mean, why not just have this exact circuit but without the 334?
Because the voltage there needs accuracy. It could be achieved with a zener but it would be noisy. With the 334 it's easily adjustable.
 
So dig this…

after reading the data sheet for the 334 for the 334th time, I’m starting to think that it’s actually got something to do with the attack or release times.

As you turn up the more knob, you allow more of the current into the system that is created by the 334 - which also happens to have a low slew rate…

Turning the knob away from More, (less?) you’re drawing from …basically R16…
I’m thinking that the reason for the inverting amp U5A is to create an attractive place for electrons, then forcing them to go through D3.

Maybe turning up the More knob just creates a higher affinity for current to be applied to a transient, which has an inverse relationship with the vca…
and maybe that said current comes from the 334, and to keep it from just chugging current straight away, the 334’s lousy slew rate acts like kind of a grouchy old gate keeper.

then the choke on the signal would open like wwwwwwooooooOOOWWWW instead of like wwWWWOWWWW

But honestly, what do I know.
I’m drinking Guinness and reading data sheets on Saturday night haha.
 
So dig this…

after reading the data sheet for the 334 for the 334th time, I’m starting to think that it’s actually got something to do with the attack or release times.
No.
As you turn up the more knob, you allow more of the current into the system that is created by the 334 - which also happens to have a low slew rate…
Current in the 334 is constant by definition.
Turning the knob away from More, (less?) you’re drawing from …basically R16…
I’m thinking that the reason for the inverting amp U5A is to create an attractive place for electrons, then forcing them to go through D3.

Maybe turning up the More knob just creates a higher affinity for current to be applied to a transient, which has an inverse relationship with the vca…
and maybe that said current comes from the 334, and to keep it from just chugging current straight away, the 334’s lousy slew rate acts like kind of a grouchy old gate keeper.

then the choke on the signal would open like wwwwwwooooooOOOWWWW instead of like wwWWWOWWWW

But honestly, what do I know.
I’m drinking Guinness and reading data sheets on Saturday night haha.
You're overcomplicating things.
 
In this circuit, the LM334 (by virtue of R23) is generates a constant current of 120 µA. As this current flows in R20, it creates a voltage drop of +1.2 V, which is buffered by U3B to become +VP. Because U5A is a unity gain inverter (R20 = R19) the output of U5A sits at -1.2 V. So the circuit generates a stable, symmetrical ±1.2 V supply. This supply, is used along with the temperature-tracking +0.6 V reference generated by U3D, R24, and D3, to process the output of RMS to DC converter U2B. It is all some elegant circuit design indeed!
 
I should point out that the voltages I talk about are with respect to an "artificial ground" of about 1/2 of the +9 V supply (see U3C). And I don't believe the choice of 9 V for the supply was arbitrary, but because this device could be powered by a 9 V battery. It's voltage would obviously drop as it discharges. But the control voltages of the RMS converter and VCA must be stable, hence the VP regulated sources. The rest of the signal processing circuitry will not care much if the battery is only 8 V - only a little loss of headroom. Again, this is some very good analog circuit design work! I like the "elegance" of getting a lot of function from minimal parts, cleverly used.
 
In this circuit, the LM334 (by virtue of R23) is generates a constant current of 120 µA. As this current flows in R20, it creates a voltage drop of +1.2 V, which is buffered by U3B to become +VP. Because U5A is a unity gain inverter (R20 = R19) the output of U5A sits at -1.2 V. So the circuit generates a stable, symmetrical ±1.2 V supply.

That is a slick circuit. Especially if you already have a current source / sink network. This would be a good way to generate +-5V for an ADC or DAC or some such.

And LM334 is an interesting part. It's a floating current source. And perhaps more important, the TO-92 is available on mouser for under $1. I'll have to remember that one.

And TL431 is another part you posted about that I noticed and must try to remember. I don't know why I'd need these yet. But I'll know why when I know why.
 
The TL431 is an amazing part that I've used often in my designs since the early 90's. Since it's basically a 2.5 V reference and an op-amp with some hefty sinking capability, with 2 resistors you can create a "super-zener" with a very sharp knee (real zeners under 8 or 9 V are really soft). The dynamic impedance is well under an ohm, making them great at supply rail decoupling.
 
In this circuit, the LM334 (by virtue of R23) is generates a constant current of 120 µA. As this current flows in R20, it creates a voltage drop of +1.2 V, which is buffered by U3B to become +VP. Because U5A is a unity gain inverter (R20 = R19) the output of U5A sits at -1.2 V. So the circuit generates a stable, symmetrical ±1.2 V supply. This supply, is used along with the temperature-tracking +0.6 V reference generated by U3D, R24, and D3, to process the output of RMS to DC converter U2B. It is all some elegant circuit design indeed!
Wow, thats heavy man! There's a lot for me to unpack here.

I was outstandingly incorrect haha I wish I could see electronics through your eyes. Thanks for clearing this up! Now I have a good foothold to understand it more.

Temperature tracking circuitry is something of a mystery to me. I haven't even attempted to read into it yet, but it's becoming up quite often in my personal studies, so maybe this is a sign to dig into it. In this case, it holds +0.6V because the diode forward voltage isn't particularly affected by temperature, is that correct?
 
That is a slick circuit. Especially if you already have a current source / sink network. This would be a good way to generate +-5V for an ADC or DAC or some such.

And LM334 is an interesting part. It's a floating current source. And perhaps more important, the TO-92 is available on mouser for under $1. I'll have to remember that one.

And TL431 is another part you posted about that I noticed and must try to remember. I don't know why I'd need these yet. But I'll know why when I know why.
I just ordered a bunch lol I'm gonna mess around and see what happens.
I wish I saw this post before I clicked 'order', because it seems like I should have thrown a few TL431 in there.
 
Thanks for the kind words AVA! Because relationship between forward voltage and forward current in a silicon diode (or transistor base-emitter junction) is exponential are exponentially related (a sort of Richter scale, where every 60 mV change in voltage changes current by a factor of 10 times or 20 dB higher). It's why the VCA control voltage and the RMS detector outputs are small voltages and are scaled in mV per dB. But, as you probably know, forward voltage of a diode also changes with temperature - at a very predictable 2 mv per degree C. So, to keep gain from changing with temperature, a voltage corresponding to temperature (in this case at the output of U3D is added to the control voltages to keep gain stable with temperature. Again, this gets to the heart of design ingenuity. I've found analog circuit design totally fascinating for just about all my adult life. If you like it, I encourage you to learn as much as you can ... it keeps getting better and better. When I was young, before college, one of the ways I learned a lot was to analyze someone else's circuit design - the designs by THAT are excellent for the purpose. But I'm dismayed at how many "cut and paste" designers are doing work for well-known companies and how many idiotic designs are offered up by self-appointed "engineers" on the internet :(. Good analog circuit designers are a vanishing breed!
 
Thanks for the kind words AVA! Because relationship between forward voltage and forward current in a silicon diode (or transistor base-emitter junction) is exponential are exponentially related (a sort of Richter scale, where every 60 mV change in voltage changes current by a factor of 10 times or 20 dB higher). It's why the VCA control voltage and the RMS detector outputs are small voltages and are scaled in mV per dB. But, as you probably know, forward voltage of a diode also changes with temperature - at a very predictable 2 mv per degree C. So, to keep gain from changing with temperature, a voltage corresponding to temperature (in this case at the output of U3D is added to the control voltages to keep gain stable with temperature. Again, this gets to the heart of design ingenuity. I've found analog circuit design totally fascinating for just about all my adult life. If you like it, I encourage you to learn as much as you can ... it keeps getting better and better. When I was young, before college, one of the ways I learned a lot was to analyze someone else's circuit design - the designs by THAT are excellent for the purpose. But I'm dismayed at how many "cut and paste" designers are doing work for well-known companies and how many idiotic designs are offered up by self-appointed "engineers" on the internet :(. Good analog circuit designers are a vanishing breed!
Thats brilliant. How can anyone come up with this stuff? I feel like there must have been a lifetime of frustration spent on finding the solution to that one temperature variance problem.

Thank you for shedding some light on it for me. I have lots to learn! I've heard other people in forums talking about the loads of unproven schematics floating around the internet. I think there's even YouTube channels dedicated to it haha. That's why when I stumbled onto THATCorp schematics it felt like a safe sandbox to really dig deep.

The internet...while good in a lot of ways, has democratized just about everything - including entitlement! Anyone can claim to be an engineer.

I'm glad I found this forum though. I've been using analog audio gear my entire life, not knowing how brilliant it really is. But I see the digital stuff creeping up, and so learning about the analog audio circuits is my way of keeping it alive before anyone too important does the cost analysis on it haha.

Cheers man, as always, I really appreciate the time you're taking to help me out.
 
This may veer off topic a little bit, but it i'll risk it for the sake of some those knowledge tidbits you guys kindly drop on me haha.
'Thought cookie crumbs' if you will...

It occurs to me, that if I we're to want to make this circuit into a 'stereo' or two channel compressor, I wouldn't actually need more than one of these RMS circuits, meaning, 5 opamps and an LM334 wouldn't need to be copied for the second channel because they would share this circuit on circuit (topology?), is that correct?

I understand two things about this:

1, I cannot simply sum the two channels together before the RMS detector, because that would give me the doubling effect and probably just add the kind of low frequencies I don't want, and false-trigger the compressor....mmm thats correct, right? It makes sense in my brain ha. Otherwise, why would it be any different than any other time you sum a stereo signal?
I got a good smack for doing this once when I was a young kid working on one of my first live shows, when I knew nothing about anything.

2, If the channels do not compress at the same time and rate, it stands to reason that the 'center' of my stereo image would bounce all over the place and probably cause a seizure or two...


THATCorp has touched on this topic here, and kind of confirms my suspicion (using the THAT2252): http://www.thatcorp.com/datashts/dn116.pdf

Aside from that, I cannot find any clear cut, straight, reliable answers on the topic. Some other forums talk about it... "just connect pins 5 and pins 15 together"..but that doesn't explain anything, seems incorrect, isn't confirmed by anything THATCorp has ever said...and probably wouldn't work for my situation. Unless I'm missing something big.

So, once again, my inexperience has once again left me with a few questions...

- In figure 5, they show Pin 6 (CAP) on both RMS detectors are connected together. Great. But the output of U1 goes...nowhere? Just floats? ...I suppose this connection would be the same for two 4315's as well.

- From what I can glean, If I was adapting this to work with the 4315 one knob circuit (from the original post), I just connect the 'CT' pins together, and bada bing. Done...lol no way it's that simple.

- In figure 6, they kind of lose me. I see nothing stereo about this schematic...It seems like that schematic needs its own design note haha.

Again, thank you for any time you put into helping me with this. You guys are gold! Maybe one day my areas of expertise will help you in return, lets hope.

- Alex
 

Attachments

  • Figure 5.png
    Figure 5.png
    361.1 KB · Views: 32
  • Figure 6.png
    Figure 6.png
    479.3 KB · Views: 29
It occurs to me, that if I we're to want to make this circuit into a 'stereo' or two channel compressor, I wouldn't actually need more than one of these RMS circuits, meaning, 5 opamps and an LM334 wouldn't need to be copied for the second channel because they would share this circuit on circuit (topology?), is that correct?
You would need two rms detectors conneced to a single threshold/volume/compression circuit.
I understand two things about this:

1, I cannot simply sum the two channels together before the RMS detector, because that would give me the doubling effect and probably just add the kind of low frequencies I don't want, and false-trigger the compressor....mmm thats correct, right? It makes sense in my brain ha.
Correct.
2, If the channels do not compress at the same time and rate, it stands to reason that the 'center' of my stereo image would bounce all over the place and probably cause a seizure or two...
Again, correct.
Aside from that, I cannot find any clear cut, straight, reliable answers on the topic. Some other forums talk about it... "just connect pins 5 and pins 15 together"..
What pin 5 and 15? Doesn't make sense to me...
- In figure 5, they show Pin 6 (CAP) on both RMS detectors are connected together.
That's called "analog OR"; that means the highest bidder takes the stake.
Great. But the output of U1 goes...nowhere? Just floats?
Yes, don't worry anout that (no pun intended).
...I suppose this connection would be the same for two 4315's as well.
That would be pin 4 on the 4315
- From what I can glean, If I was adapting this to work with the 4315 one knob circuit (from the original post), I just connect the 'CT' pins together, and bada bing. Done...lol no way it's that simple.
Actuually yes.
- In figure 6, they kind of lose me. I see nothing stereo about this schematic..
Figure 6 of what document?
 
Back
Top